Credits and billing

One page, one standard credit

A standard page visit costs 1 credit whether it returns one product, 50 products or no usable records. Rows, products and variants are output measures, not billing units.

Ten listing pages cost 10 credits. Opening a listing and 25 product detail pages costs 26 credits. The current Shopify pilot additionally visits a currency-check page, so these examples total 11 and 27 credits; the latter exceeds its 25-page run cap.

Decide what to follow

Pagination chooses the total result pages per listing, including the first. Linked products chooses the first N detail pages across all visited listings. Collections chooses the first N other collections from one collection-directory page; that directory also consumes credit. Collections have no recursive levels in this adapter.

The adapter visits the currency check, initial listing and pagination, collection directory and collections with their pagination, then linked product details. Other page types are not followed. A homepage input selects the catalogue feed; the HTML homepage is not fetched.

The default is 2 pages and 2 credits, with following off. Increase both budgets if you want more visits. The estimate is a bounded maximum, not a promise that all selected pages exist.

Discovery and failures count

Any source visit consumes credit, including verification, discovery, navigation, empty responses and failed source requests. Inspecting links in an already processed page adds no cost. Visiting a discovered URL adds a page credit.

Credits are reserved before dispatch. The run stops before exceeding either budget and releases the unused reservation at settlement. Waiting for review or validating an input locally makes no network visit and consumes no credit. A request can finish with zero usable records and a nonzero page charge.

Identity and retries

Within a run, known store aliases, product URLs inside collection paths, tracking parameters and fragments normalize to the same adapter source. Meaningful generated page and limit parameters remain distinct. Arbitrary input filters are not applied. Redirect responses count but are not followed. HTML canonical hints are not read by this JSON adapter.

Queue retries reuse the durable page journal, including empty or failed responses. An interrupted visit is never automatically re-fetched; its reserved page credit counts conservatively and the run reports the interruption. A new run is a new authorization.

Simple difficulty rates

Only standard JSON extraction at 1 credit per page is enabled. Browser rendering, interactions and complex extraction tiers are not available. A future adapter must quote any higher page rate before it can run; record count cannot be used as a difficulty multiplier.

Progress and history

During execution, page visits and consumed credits appear alongside the reservation. Consumed credits are included in the reservation, not added to it. After completion, expand Where credits went to inspect source URLs, page types and outcomes. Historical runs retain their old settled charges and are labeled as the previous model.

Trial credits and top-ups

New accounts receive 100 trial credits once. Website and API requests share the balance. Buy additional credits through Stripe Checkout in the Billing tab. Purchases are one-time, with no automatic recharge.

US$10 buys 10,000 credits; US$25 buys 26,250 (5% bonus); US$100 buys 110,000 (10% bonus); US$600 buys 708,000 (18% bonus). Credits are added only after verified payment, once per purchase. Refunds reverse the proportional share of all purchased credits, including bonuses. If those credits were already used, the next top-up first covers the difference. Contact support with a run ID to investigate a page charge.